Predictive control of transient engine testbeds

Conventional testbed systems use a number of independent three-term or proportional integral derivative controllers. These controllers are relatively cheap and are sufficiently robust to operate on a wide variety of engines for steady-state testing. However, tuning for transient control is difficult, time-consuming and does not normally lead to satisfactory performance. This problem is accentuated in a research and development environment where many different types of engines are tested and their characteristics constantly modified. In order to maintain accurate control, the controllers must be returned for each particular engine configuration. The compromise for steady-state testing, in which the controller is deliberately detuned to have a slow but safe response, is unacceptable for transient control. It is for these reasons that researchers have investigated modern testbed control algorithms. This paper presents a state-space formulation of generalised predictive control (GPC) and compares its performance with the standard polynomial GPC algorithm when applied to testbed speed and torque control